Volume 7, Issue 1 -
Spring/Summer 1999
Volume 6, Issue 3
Fall 1998
Volume 6, Issue 2
Spring/Summer 1998
Volume 6, Issue 1
Winter 1998
Volume
5, Issue 4
Fall 1997
Volume
5, Issue 3
Summer 1997
Volume
5, Issue 2
Spring 1997
Volume
5, Issue 1
Winter 1997
Volume
4, Issue 4
Fall 1996
Volume
4, Issue 3
Summer 1996
Volume
4, Issue 2
Spring 1996
Volume
4, Issue 1
Winter 1996
Volume
3, Issue 4
Fall 1995
Volume
3, Issue 3
Summer 1995
Volume
3, Issue 2
Spring 1995
Volume
3, Issue 1
January 1995
Volume
2, Issue 4
October 1994
Volume
2, Issue 3
July 1994
Volume
2, Issue 2
April 1994
Volume
2, Issue 1
January 1994
Volume 1, Issue 4
October 1993
Volume
1, Issue 3
July 1993
Volume
1, Issue 2
April 1993
Volume
1, Issue 1
January 1993
|
High Performance Fortran At Last
Ken Kennedy, Director, CRPC
On May 15, 1993, Version 1.0 of the High Performance Fortran (HPF)
language specification will be available for public distribution. This
document is the product of the High Performance Fortran Forum, a unique
collaboration by industry, academia and government laboratories. There
were more than 40 people in the working group that met every six weeks
between January 1992 and December 1992 to produce a draft for public
comment and in March 1993 to finalize the document based on the public
review. The result specifies a language that is a blend of Fortran 90,
data distribution specifications similar to those in Fortran D, and new
parallel constructs such as the FORALL statement. Twelve companies (ACE,
APR, DEC, Intel, KAI, MasPar, Meiko, Lahey, NA Software, Portland Group,
PSR, Thinking Machines) have announced that they will offer products
based on HPF and another seven companies (Archipel, Convex, Cray
Research, Fujitsu, HP, IBM, nCube) have actively participated in the
standardization process, although they are not yet ready to announce
products.
The next year will be critical for High Performance Fortran, because its
success or failure will depend upon the effectiveness of the first
compilers in generating efficient code for various target architectures.
If it is successful, HPF could help break down one of the most
formidable obstacles to progress in the high performance computing
industry, namely the absence of a vehicle for expressing data parallel
programs in a machine-independent way.
The current version of HPF is helpful only for the regular problems,
i.e., those in which all subscript expressions are simple affine
functions of the loop induction variables. In the long term, HPF cannot
be a complete success unless it can also help with problems on irregular
or adaptive meshes (characterized by the use of subscripted variables
within subscripts). In collaboration with Joel Saltz's group at the
University of Maryland, the Center for Research on Parallel Computation
(CRPC) Fortran D groups at Rice and Syracuse have embarked on a project
to investigate how to support irregular problems in a language like HPF.
If this research bears fruit, it may influence the second round of HPF
standardization, scheduled for the 1994 calendar year.
The CRPC has played a central role in the development of HPF. It served
as a convening organization for the HPF Forum and provided support costs
for the academic participants, assuring that the views from academia
were adequately represented. In addition, the Fortran D research deeply
influenced the final language specification. I believe that this is a
perfect example of how a research center like the CRPC can have maximum
impact on industrial practice and products. The HPF process has already
served as a model for the ongoing effort to standardize message-passing
interfaces, in which the CRPC is also involved. I do not think it would
be much of an exaggeration to say that HPF is the single most important
technology transfer accomplishment by the CRPC or any other science and
technology center over the past five years. As director, I am proud of
the role the CRPC has played in this effort and I will continue to
encourage activities of this sort throughout the lifetime of the center.
Table of Contents
|